Program seminářů pro rok 2010/2011
- 23. 9. 2010
- Úvodní seminář podzimního semestru
Informace o koncepci semináře v podzimním semestru.
Domluva programu semináře.
- 30. 9. 2010
- Informace o doktorském projektu OPVK. Diskuse.
- 7.10. 2010
- Tomáš Hnilica: Visualization of large FEM models.
Numerical simulation is an important part of computer aided product development chain and FEM simulations are widely used. The amount of data, produced by a finite element calculation places a particular challenge to scientific visualization. The aim of my work is visualization of large FEM models on commodity hardware. In this presentation the overview of FEM model and simulation process will be given. Design of out-of-core visualization system will be presented.
- 14.10. 2010
- Vojtěch Bystrý: Model for non-linear local motifs recognition
Short local motifs are the basis for the protein folding prediction algorithms. These motifs are usually recognized in the protein sequence sequentially, which might not be the best way to do it. Since protein sequences have very non-linear context characteristic it is more efficient to split the sequence to different non-linear context groups. Model we propose is trying to do just that with its ability to switch between context groups and remember the point of the switch in these groups .
- 14.10. 2010
- Adam Husár: Application-Specific Instruction Set Processor Design and Programming
Presentation will consist of two parts. The first one will be an introduction to Application-Specific Instruction set Processor (ASIP) design and to related Electronic System Level design methodologies. As a case study will be shown design of an H.264 encoder/decoder processor SVENm from Austrian company OnDemand Microelectronics. The second part will focus on ASIP design Architecture Description Language ISAC and on transformation from an ISAC model to a model suitable for compiler generation. Preliminary results for automatically generated C compiler for the MIPS architecture will be shown.
- 4.11. 2010
- Filip Konečný: Relational Analysis of Non-deterministic Integer Programs
Computing transitive closures of integer relations is the key to finding precise invariants of integer programs. In this talk, we describe an efficient algorithm for computing the transitive closures of difference bounds, octagonal and finite monoid affine relations. On the theoretical side, this framework provides a common solution to the acceleration problem, for all these three classes of relations. In practice, according to our experiments, the new method performs up to four orders of magnitude better than the previous ones for the first two classes. This algorithm provides the basis for the analysis of non-deterministic integer programs (also known as counter automata or counter systems). We explain how transitive closure computation combines with other methods in order to provide a framework for the analysis of such systems. These ideas led us to the implementation of the FLATA tool. (Joint work with Marius Bozga, Radu Iosif, and Tomas Vojnar.)
- Tomáš Babiak: An improvement of the translation of LTL to Büchi automata
The standard technique for LTL model checking consists of translating the negation of the LTL specification into a Büchi automaton A, and then of checking if the product M x A has an empty language. Despite the main problem of this verification method is the state explosion problem, the translation of LTL formulae into equivalent Büchi automata itself can take a significant amount of time of the whole process. We present an improvement of an algorithm translating LTL formulae into Büchi automata via alternating automata. In particular, we improve the transformation of alternating Büchi automata to generalized Büchi automata, where we temporarily ignore some transitions leading from the alternating automata states corresponding to subformulae that are prefix-invariant. Experimental results show that our improvement can lead to faster translation.
- 4.11. 2010
- Vít Baisa: Natural Language Processing from Scratch
Natural languages are complicated. To be able to even talk about them we introduce terms as part of speech, case, semantic role, tense, hyperonym and many others. These terms are based on our observation of language but they always generalise the much more intricate reality. Any generalisation inevitably brings errors into our understanding of language. We usually process natural languages with relatively sparse data but with complicated algorithms. My aim is, on the contrary, to build a large data source representing language and to implement very simple algorithms to deal with it. In my presentation I will show a completely new view of language (especially Czech) - extremely simple and as universal as possible. I will also describe how to build the extensive data source and how to solve some tasks of natural language processing (syntactic analysis, word sense disambiguation, machine translation) with such data source and within the new view. The view of language is very simple and does not require any previous linguistic knowledge. That is why you are sincerely welcome to join subsequent discussion. Since my work has started from scratch and it is still under development, I will be grateful for your suggestions and comments.
- 25.11. 2010
- Jan Samek: Modeling Multiple-Context Trust in Multi-Agents systems
Artificial agents are autonomous entities that should behave rationally by their own control. They should have also some social abilities that enable them to make strategic decision inside a multi-agent community. One of such decision is to select proper partner which it believes that its behaviour will be reliable and trustworthy. We address area of trust which may be important for improving agent's reasoning capabilities and by this also rationality of its behaviour within an agent or multi-agent system. Presentation introduces specification of the model which is used for representation of multi-context trust. This model is in fact a part of agent’s belief base and it is used when decision about some system element trustworthy takes part in agent's reasoning process. For these purposes, we develop Hierarchical Model of Trust in Context which is represented by multilevel graph, where each node of the graph represents different aspects (contexts) of trustee and each edge of the graph represents relationship between different aspects. This is useful to modelling trust in respect to multi-context environment.
- 2.12. 2010
- Michail Tsikirdekis: Aggression in Social Networks
Aggression in social networking websites is one of the behaviours that people exhibit online that could destabilize an otherwise stable community. Research from the fields of psychology and sociology shows that there some subtle differences between the ways that people express violence online and in real life, usually because of the restricting nature of the computer-mediated communication. The rest of the core elements of violent behaviour though remain the same and are well known facts. One of the top theories of aggression is the frustration-aggression theory which can easily be observed online every day. Most of the factors that contribute to frustration and in effect aggression are part of the digital environment which is artificially created by the software engineers. This environment due to its artificial nature can be manipulated, altered, or changed. Software engineers need to learn how to alter the user interface in order to achieve the desired effects on their social networking software. In depth analysis and research should be conducted in order to find how the architectural design of application affects the interactions of the people that use the software.
- 2.12. 2010
- Vojtěch Zavřel: Similarity based event-detection system for videos
With the growing number of videos the searching for an appropriate event in video datasets is becoming more and more complicated. Many of currently available systems are able to search according to one of the video aspects only, are domain specific, or have problems with performance. Our aim is to build a video event-detection system based on similarity principles that are generalized by the abstraction of the metric spaces. This framework will be domain nonspecific; it will be able to combine several video aspects based on a user-supplied aggregation function; it will not require any learning phase; and it will be capable of processing the volume of data significantly higher than currently available systems.
- Jaroslav Bayer: Towards Text Mining in Technology-Enhanced Learning
In technology-enhanced learning environment, text mining can be useful for clustering course information, classification of student comments, for theses classification and evaluation, and others. We describe a pre-processing step and a method of text classification. The best results have been obtained for classification of theses. We discuss results obtained for classification of these three kind of texts and explain what additional information need to be exploited for improving a classification accuracy.
- 16.12. 2009
- Poster Session
Prezentovány budou postery vytvořené v podzimínm semestru.
- 19. 2. 2011
- Úvodní seminář jarního semestru
- Informace o koncepci semináře v jarním semestru. Domluva programu semináře. Diskuse.
- 3. 3. 2011
- David Sehnal: Analyzing 3D Structural Motifs in Proteins
Nowadays, a large amount of information about the 3D structure of proteins and other types of molecules is available in databases such as the Protein Data Bank. This allows us to analyze biochemically important motifs in these molecules. We present several methods of analysis including superimposition, EEM charge calculation, and computing tunnels in proteins. Furthermore, the emphasis is on composing these analyzes. This is achieved through software called WebChemistry - a user friendly web-based interface for manipulation with chemical data.
- 10. 3. 2011
- Sven Dražan: Analysis of biological dynamic reaction systems
At the basis of all living organisms are cells. Their behaviour is caused by intricate systems of reacting molecules such as large proteins, enzymes, signaling molecules and metabolites. Depending on the amount of knowledge and scale of interest different abstractions are used to express the dynamic behaviour of such systems. If the species concentration changes are given in terms of ordinary differential equations and a method for numerical simulation is used to approximate the time evolution of such a system a single trajectory from the initial conditions point can be obtained. Combined with methods from model checking a property can be evaluated to hold or not over a trace taken from a single simulated trajectory. The presented approach will show possibilities to infer about the validity of properties for whole sets of trajectories starting in parts of a given region of initial conditions.
- 17. 3. 2011
- Moslem Amiri: Evaluation of Lifetime Bounds of Wireless Sensor Networks
In this presentation, we estimate lifetime bounds of a network of motes which communicate with each other using IEEE 802.15.4 standard. Different frame structures of IEEE 802.15.4 along with CSMA/CA medium access mechanism are investigated to discover the overhead of channel acquisition, header and footer of data frame, and transfer reliability during packet transmission. This overhead makes the fixed component, and the data payload makes the incremental component of a linear equation to estimate the power consumed during every packet transmission. Finally we input this per-packet power consumption in a mathematical model which estimates the lower and upper bounds of routings in the network. We also implemented a series of measurements on CC2420 radio used in a wide range of sensor motes to find the fixed and incremental components, and finally the lifetime of a network composed of the motes using this radio.
- 24. 3. 2011
- Jaromír Plhák: Processing of Graphics for the Visually Impaired
In this presentation, current approaches to graphics processing for the visually impaired are introduced. It involves a graphics exploration as well as a picture generation by the visually impaired and blind. Our dialogue based approach that integrates graphical ontologies, what-where language and recursive navigation grid is described in more detail. Well known graphic processing applications for the users with visual impairment are demonstrated as well.
- 31. 3. 2011
- Jaroslav Bayer: Retrieval and analysis of reports on a fire brigade operations
The presentation deals with the searching of texts on the internet based on the information obtained from the firemen database. The goal is to classify texts into several categories. We describe the methods of text collection and text mining. The highest overall classification accuracy – by the type of incident and geographic situation – reaches 86 % and 96 %. The given database consists of more than 116,000 records. We have tried to search and download texts describing about 4,000 selected events. We have succeeded in more than 2,900 cases.
- 7. 4. 2011
- Šimon Suchomel: Finding plagiarism by searching the Internet
Plagiarism detection has become even more important since variety of documents are electronically available. In the IS MU there is a tool for discovering plagiarism by evaluating document similarities among all documents stored in the IS MU and other systems with IS MU interconnected. For discovering similarity of newly added document, the system must contain the source or the referenced document which was potentially copied. One way is to provide the system a relevant document or set of documents which may be used as the source of plagiarism. This presentation deals with design of system for automated discovering and processing of similar documents in the Internet. It also discuss other benefits of input document processing and future improvements to IS MU plagiarism detection system.
- 14. 4. 2011
- Michail Tsikerdekis: Aggression and anonymity
The effects of anonymity on aggression have been discussed by many social scientists in the past years. Anonymity is a factor that could lead to dis- inhibited behavior which could cause damage to many online communities. This knowledge provides software engineers with a dilemma as to whether to use anonymity as an option for their users and suffer the increased aggression that their community might exhibit or exclude it all together. Anonymity is a generic term and should be analyzed further into different states such as pseudonymity and complete anonymity. There is no knowledge covering the effects of these two different states of anonymity in the interactions of users in relevance to aggression. In this paper an experiment was conducted in order to determine if there is any effect of the two anonymity states to the users interactions and how aggressively they respond to certain situations, and also how they differ from each other. The findings show that in general, there is no difference in how aggressive a user responds between pseudonymity, com- plete anonymity or when users answer with their real names. However when the differences were tested in the context of how strongly opinionated the users were about a topic, users with extremely strong opinions were found to respond more aggressively under the state of pseudonymity.
- 21. 4. 2011
- Jana Fabriková: A Discrete Abstraction of Biological Systems
Our aim is to find a suitable discrete abstraction of a biological system. By a biological system we mean a finite set of biochemical species together with a set of biochemical reactions transformed into a set of ordinary differential equations in the concentrations of species. By a suitable discrete abstraction we mean an automaton with finite number of states, each of which can be described by a bounded set of integer numbers. There are several existing approaches to this problem that will be briefly introduced and our approach (a combination of discretization of the phase space with local numerical simulations) together with its advantages and disadvantages will be described.
- 28. 4. 2011
- Fedor Tiršel: On ontology-based indoor navigation for the visually impaired
Blind and visual impaired face many problems during their daily activities with navigation, especially in environments unknown or not designed with assistive purposes. Many systems try to cover navigation issues of handicapped users. The presentation overviews the ontology-based approach of indoor navigation. The developed system consists of RFID tags, portable reader which can be integrated into the long cane and smart phone. Technologies provided by smart phones have the potential to support future needs of the system.
- 12. 5. 2011
- Lukáš Kalčok: Haptic Rendering and Algorithms
During the last decade of computer graphics there has been a major progress in the field of haptic rendering. Nowadays haptic rendering is used in many fields such as molecular docking or virtual sculpting. The presentation will include introduction to the field, algorithms used in haptic rendering, used equipment and solutions used in laboratory of Human Computer Interaction at Faculty of Informatics at Masaryk University
- 19.5. 2011
- Poster Session
Prezentovány budou postery vytvořrné v jarním semestru.